Strong Enhancement of the Radiative Decay Rate of Emitters by

agreement with electrodynamic model calculations that predict a maximum attainable efficiency around 80% due to nonradiative losses in the metal. In r...
0 downloads 10 Views 1MB Size
NANO LETTERS

Strong Enhancement of the Radiative Decay Rate of Emitters by Single Plasmonic Nanoantennas

2007 Vol. 7, No. 9 2871-2875

O. L. Muskens,*,† V. Giannini,‡ J. A. Sa´nchez-Gil,‡ and J. Go´mez Rivas† FOM Institute for Atomic and Molecular Physics AMOLF, c/o Philips Research Laboratories, High Tech Campus 4, 5656 AE EindhoVen, The Netherlands, and Instituto de Estructura de la Materia, Consejo Superior de InVestigaciones Cientı´ficas, Serrano 121, 28006 Madrid, Spain Received July 2, 2007; Revised Manuscript Received July 26, 2007

ABSTRACT We demonstrate a strong, 5-fold enhancement of the radiative decay rate from highly efficient fluorescent dye molecules around resonant optical nanoantennas. The plasmonic modes of individual gold dimer antennas are tuned by the particle length and the antenna gap, providing control over both the spectral resonance position and the near-field mode profile of the nanoantenna. Resonant enhancement of the radiative and nonradiative decay rates of a fluorescent dye is observed, resulting in an increase of the internal quantum efficiency from 40% up to 53% for single antennas, and up to 59% for antenna clusters. This improvement of the already high quantum efficiency of the dye molecules is in agreement with electrodynamic model calculations that predict a maximum attainable efficiency around 80% due to nonradiative losses in the metal.

In recent years, there has been an enormous interest in plasmonic nanostructures for applications in biosensing, nanophotonics, and enhanced light emission.1,2 Surface plasmons give rise to large local field enhancements, which may be used for nonlinear optical spectroscopy such as SERS, multiphoton absorption, and for enhancement of fluorescence. Although many studies have addressed the enhancement of fluorescence, it has turned out to be extremely difficult to selectively enhance only the radiative decay channels using plasmons. Selective modification of only radiative decay channels is crucial for light-emitting devices that benefit only from radiative enhancement. The further improvement of highly efficient emitters with nanostructures in general, and specifically with plasmonic nanostructures, represents a scientific challenge with important technological consequences. For rough metal films, extremely large (up to 103) increases in intrinsic relaxation rates were reported, which were however governed by nonradiative losses in the metal.3,4 Recent investigations on small metal nanoparticles5-8 have led to significant advancements in the understanding of the effect of emitter-particle distance on the fluorescence yield. Various other geometries have been under investigation recently, including nanoshells9 and lithographic particle * Corresponding author. E-mail: [email protected]. † FOM Institute for Atomic and Molecular Physics AMOLF. ‡ Instituto de Estructura de la Materia, Consejo Superior de Investigaciones Cientı´ficas. 10.1021/nl0715847 CCC: $37.00 Published on Web 08/07/2007

© 2007 American Chemical Society

arrays with various shapes and sizes.10,11 Although in many cases an increase of the total fluorescence intensity has been observed, a large portion of this has to be attributed to enhancement of excitation rates. The few reports in which effects on the excitation and spontaneous emission could be unambiguously separated experimentally3,7,11 showed a predominant enhancement of nonradiative decay rates and hence a quenching of fluorescence emission. From the above investigations, the crucial picture emerges that the most critical step in plasmon-enhanced fluorescence emission is not the coupling of the emitter to plasmonic modes, but rather the efficient radiative outcoupling of surface plasmons into photons. We investigate here a novel class of nanoparticle designs that are optimized specifically with respect to this radiative outcoupling of plasmons into far-field light. These so-called optical antennas have recently been proposed based on analogies with radiowave antennas12,13 and combine a high radiative efficiency with strong geometric resonances and large local field enhancement in the gap between two interacting nanoparticles. This strong improvement of the matching of far-field optical radiation with local dipoles makes optical antennas extremely promising elements for nonlinear spectroscopy12,13 and for extraction of light from emitters.14,15 To conclusively identify the different contributions to the plasmon-enhanced fluorescence, we combine in our experiment several key elements, namely: (i) precise variation of

relevant antenna parameters, (ii) detailed knowledge of the plasmonic mode spectrum, (iii) separation between plasmonic effects in the excitation and emission rates, and (iv) extraction of intensities and decay rates using fluorescence lifetime imaging. In contrast to previous reports in which low quantum-yield emitters were used,9,10 we choose as starting point a dye with a relatively high quantum efficieny of 40% (Atto680, Atto Tec GmbH). The large Stokes shift of ∼70 nm of this dye allows a separate treatment of excitation and emission rate effects in the fluorescence. We specifically consider dimer antennas, consisting of two rod-shaped arms, because these are the simplest realization of a rich diversity of potential nanoantenna structures. Several experimental and theoretical investigations have shown that extended-size nanorods act as standing-wave resonators for surface plasmon polaritons, exhibiting geometrical halfwavelength resonances with spectral positions depending strongly on antenna arm length.13,16-18 We unambiguously show here that the coupling between the two antenna arms results in an improved average response of the antennaemitter complex compared to resonant antennas with uncoupled arms. We find that, while the linear scattering spectra are very similar, the decay rate enhancements show a pronounced difference, which clearly demonstrates the importance of the near-field mode profile besides the spectral resonance structure. We also present rigorous calculations of the decay rate of a dipole in the proximity of the nanoantenna, which are in qualitative agreement with the measurements. These calculations predict a limit to the enhancement related to nonradiative losses in the metal. Arrays consisting of individual gold nanoantennas of 20 nm thickness were fabricated using e-beam lithography.19 Subsequently, a 10 ( 2 nm silica spacer layer was deposited, followed by an active layer of 10 nm thickness fabricated by spin-coating of a concentration (3 × 10-5 M) of dye molecules embedded in a polymer (PVB) matrix. The selected dye concentration resulted in a coverage of several hundred dye molecules per µm2. The highly photostable dye has a fluorescence decay time of 3.3 ns with a corresponding quantum efficiency η of 40% in ethanol and emits in a spectral band around 730 nm. A scanning electron microscopy (SEM) image of a dimer nanoantenna consisting of two 90 × 60 × 20 nm3 antenna arms with an antenna gap, ∆, of 20 nm is shown as an inset of Figure 1. The fluorescence lifetimes of dye molecules deposited around single nanoantennas were measured using a timecorrelated single-photon counting setup20 (Picoquant GmbH) synchronized with a laser-scanning confocal microscope (Nikon C1). In this setup, the fluorescence decay as a function of time was measured while spatially scanning the excitation laser beam over the sample. The molecules were excited by light from pulsed diode laser with 150 ps pulse duration, polarized parallel to the antenna long axis. The pump beam was focused to a near-diffraction limited spot of less than 0.5 µm by means of a 100×, 0.9 NA, microscope objective, resulting in an energy per pulse of ∼50 µJ/cm2 on the sample. Note that the excited volume is still considerably larger than a single nanoantenna. Emitted 2872

Figure 1. Normalized fluorescence decay measured at a resonant nanoantenna (diamonds, black) and away from the nanoantenna (open dots, red), with exponential decay fits (lines, black). Inset: SEM image of the resonant gold nanoantenna, consisting of two 90 × 60 × 20 nm3 gold nanorods with an antenna gap of 20 nm.

fluorescence photons were collected using the same 0.9 NA microscope objective and detected by means of a photoncounting avalanche diode. Figure 1 shows fluorescence decay curves taken at the position of a resonant nanoantenna (diamonds, black) and more than 1 µm away from the antenna (circles, red). In the absence of an antenna, the decay is single-exponential with a decay rate, γ, identical to the value measured in ethanol, γ0 ) 0.3 ns-1. At the position of the antenna, the fluorescence decay is multiexponential with a slow component, γ1, equal to γ0. We attribute this slow component, indicated by the red area in Figure 1, to a background of dye molecules in the excited volume that do not couple to the antenna. The fast component can be fitted to an exponential with a decay rate, γ2, of 1.7 ns-1, more than five times larger than γ0. We assign this contribution, indicated by the blue area in Figure 1, to a fraction of molecules that is coupled to the nanoantenna. A complete spatial and temporal image of fluorescence decay around single nanoantennas was obtained from twodimensional scans in which time-resolved fluorescence decay curves like those in Figure 1 were measured at several positions. Typical results for a resonant antenna are presented in Figure 2. Figure 2a shows an image of the total fluorescence intensity, Itot, emitted with a polarization parallel to the nanoantenna long axis, normalized to the average intensity, I0, far away from any antenna. We observe an increase of the intensity to 1.3 times I0 at the position of the nanoantenna. For polarization of the emitted light perpendicular to the antenna, shown in Figure 2b, no increase is observed. Because the polarization of the excitation light remained fixed, this indicates that the increase of the fluorescence intensity is related to the emission properties of the antenna-coupled dye fraction and not just to an enhancement of the pump intensity. The corresponding timeresolved fluorescence curves have been fitted to a doubleexponential decay function I(t) ) A1 e-γ1t + A2 e-γ2t

(1)

yielding the amplitudes A1, A2 and decay rates γ1, γ2 for the Nano Lett., Vol. 7, No. 9, 2007

Figure 2. (a) Fluorescence intensity around a single resonant gold nanoantenna [on-scale SEM image in inset of (a)], for polarization of the detected light parallel, and (b) perpendicular to the antenna axis. Scale bars denote 1 µm. (c,d) Double-exponential fitted amplitudes for slow and fast decay components A1 and A2, normalized to value A0 in absence of antenna, for parallel emission polarization. (e) Horizontal cross-sections of Figure 2c,d. (f) Fast decay component γ2 normalized to γ0.

slow and fast decay components, respectively. The amplitudes are shown in Figure 2c,d, normalized to the singleexponential background amplitude in absence of the antenna, denoted by A0. The increase of A2 and reduction of A1 around the nanoantenna are also represented in Figure 2e, where a horizontal line scan of Figure 2c,d across the middle of the nanoantenna is shown. The fast component A2 increases to three times the background A0 within a near-diffraction limited spot with a diameter of ∼0.5 µm. The slow component A1 decreases to around 0.8 A0, indicating that only a relatively small fraction of molecules (∼20%) in the diffraction-limited spot couples to the nanoantenna. The fast decay rate, γ2, shown in Figure 2f, is only present around the nanoantenna, where we observe a 5-fold reduction of the lifetime. From the experimental values of the intensities and fluorescence decay rates, we can obtain the corresponding radiative and nonradiative components by making one justified assumption, namely we estimate an average quantum efficiency enhancement η/η0 from the local intensity ratio Itot/I0. By doing this, we assume that no contributions are present from pump field enhancement. This assumption is justified by the polarization-dependence of the intensity enhancement in Figure 2a,b and by the low, nonresonant scattering at the pump wavelength, as observed in Figure 3a,b (diamonds, green), as will be discussed below. The definition of the quantum efficiency in terms of the ratio of the radiative to the total decay rate η ) γR/γtot can be inverted, yielding γR ) ηγtot and γNR ) (1 - η)γtot. For the Nano Lett., Vol. 7, No. 9, 2007

Figure 3. Analysis of coupled (a,c,e) and uncoupled (b,d,f) nanoantennas, respectively, with antenna gap widths of ∆ ) 20 nm and ∆ > 80 nm, against antenna arm length L. (a,b) Scattering strength at the excitation (λ ) 660 nm, diamonds, green) and emission (λ ) 730 nm, circles, red) wavelengths for longitudinal polarization. Lines denote calculated cross sections using model calculations (see text).19 (c,d) Enhancement of the radiative (open dots, black) and nonradiative (diamonds, black) rates γR and γNR. (e,f) Average quantum efficiency. Lines are guides to the eye.

dye molecules coupled to the antenna, the experimental total decay rate γtot equals γ2. In the final analysis, each parameter is averaged over eight identically designed antennas to reduce variations due to the local distribution and orientations of dye molecules. We measured the fluorescence decay for single nanoantennas with various arm lengths, L, for the two cases of strongly coupled arms (gap width ∆ ) 20 nm) and virtually uncoupled arms (∆ > 80 nm). Parts c and d of Figure 3 show the resulting values of γR (open circles) and γNR (closed diamonds), relative to their values in absence of the metal structures, for the two cases of strongly coupled (c) and uncoupled (d) antenna arms. For the strongly coupled antenna arms (∆ ) 20 nm), a maximum of enhancement occurs for the shortest antennas (L ) 80 nm). The combined resonant enhancement of both γR and γNR results in an increase of the average quantum efficiency from 40% up to 53 ( 3% for the resonant antennas with small gap widths, as shown in Figure 3e (circles, blue). For the uncoupled antennas (∆ > 80 nm), the enhancement of both decay rates and of the quantum efficiency is less strong than those for the coupled antennas (Figure 3f). The results of Figure 3 demonstrate that resonant dimer antennas with strongly coupled arms yield a larger enhancement of spontaneous emission than uncoupled antennas. 2873

To relate the observed enhancements to the plasmonic mode structure of the nanoantennas, we have also investigated their spectral response. The antennas are illuminated by a broad-band source using a dark-field microscope objective, while the far-field scattered intensity is detected by an avalanche photodiode. A more detailed spectroscopic study of the individual nanoantennas is published elsewhere.19 A clear longitudinal plasmon resonance occurs for the individual nanoantenna arms that is hybridized for small gap distances when the arms are capacitively coupled.16,21 In parts a and b of Figure 3, we present measurements of the antenna scattering strengths in the two relevant spectral regions of the fluorescence measurements, namely the excitation wavelength (660 nm, green diamonds) and emission band of the dye (730 nm ( 30 nm, red circles). The intensity in the emission band clearly shows a resonant dependence on L, corresponding to the shift of the longitudinal surface plasmon mode through the selected spectral window. For the strongly coupled arms (∆ ) 20 nm), this resonance is red-shifted with respect to the individual particle resonance, leading to a maximum at smaller lengths L.19 The scattered intensities taken at the excitation wavelength (green diamonds in Figure 3a,b) do not show such a resonant behavior because these are located next to the longitudinal plasmon mode. Clearly, both the radiative and nonradiative rates (Figure 3c,d) and the average quantum efficiencies (Figure 3e,f) show a strong correlation with the scattering strength in the emission band. The absence of enhancement for the shortest antennas with ∆ > 80 nm validates the assumption of negligible excitation field enhancement. To model the fluorescence enhancement, we have made use of the rigorous scattering equations of the Green’s theorem surface integral equations to calculate the local nearfield intensity in the corresponding two-dimensional geometries.19,23 The result of these calculations are shown in the upper images of Figure 4. For the strongly coupled antenna arm (∆ ) 20 nm), a strong enhancement of the local intensity of up to 102 is found in the center of the gap. For extended antenna sizes as those of this study, it has been shown experimentally17,18 and theoretically13,16 that the longitudinal resonances correspond to a half-wave dipolar mode. This mode profile is considerably different from that of the small spherical particles considered in classical electromagnetic models for fluorescence enhancement.22 Next, the scattering formulation is exploited to determine the modification of spontaneous emission of dye molecules by the local plasmonic mode density. This is done by treating them as classical dipoles p, where the total decay rate, γ ) γNR,0 + γNR + γR (where γNR,0 is the intrinsic nonradiative decay rate of the dye), is obtained from the imaginary part of the local, self-consistent electric field Elocal, through24 γ/γ0 - 1 ∝ (η0/|p|2)Im{p*‚Elocal}

(2)

Elocal is calculated from the rigorous scattering formulation; radiative and nonradiative decay rates are discerned by performing the calculations with and without metal absorption. In parts a and b of Figure 4, we present results for a 2874

Figure 4. Modified spontaneous emission calculated for a dipole emitter at 730 nm wavelength, positioned at the location of maximum coupling to the antenna and oriented parallel to the antenna axis. Radiative and nonradiative decay rates γR, γNR are shown vs antenna arm length L for the two cases of (a) strongly coupled antenna arms (∆ ) 20 nm) and (b) uncoupled antenna arms (∆ ) 80 nm). (Top graphs) Calculated near-field intensities around two resonant antennas for plane-wave illumination,19 with spots (cyan) indicating the location of the emitter in the calculations (scale bar denotes 100 nm).

dipole positioned at the location in the gap where the largest modification of the spontaneous emission takes place (indicated by a dot in the near-field maps of Figure 4). Clearly, both the calculated radiative and nonradiative rates in Figure 4a,b show the same strong dependence on L, as seen in the experiments. The calculated enhancements are a factor of 3 higher than the experimental ones and show a larger difference between radiative and nonradiative components. These differences are due to the fact that the calculations are performed at the point of the maximum field enhancement, while in the experiments, we measured the average decay rate close to the antenna. Also, in the experimental analysis, we used an ensemble-averaged quantum efficiency that overestimates the nonradiative part in the total decay rate. A complete ensemble calculation would involve a complex averaging over all positions and orientations of molecules, which is beyond the scope of this work. The positive influence of antenna arm coupling on the total decay rate, observed both experimentally and theoretically, implies a direct modification of the local field enhancement. Our model allows us to attribute this to the formation of a “hot spot” in the antenna gap and edges, as seen in the corresponding near-field profile of Figure 4, resulting in an increase of the local density of plasmon modes. The combination of the radiative and nonradiative decay rate enhancements of Figure 4a with the experimental efficiency of the dye results in a maximum achievable internal quantum efficiency of 80%, limited by the intrinsic nonradiative losses in the metal. To explore the limits of the maximum achievable quantum efficiency with nanorod dimers as presented above, we have increased the fraction of antenna-coupled dye molecules by fabricating clusters of antennas containing up to 4 dimers within a diffraction limited spot size. The resulting average quantum efficiency enhancements for the most resonant Nano Lett., Vol. 7, No. 9, 2007

FIS2006-07894 and FIS2004-0108) and Communidad de Madrid through the MICROSERES network (grant S-0505/ TIC-0191) and V.G.’s Ph.D. scholarship. This work is part of the research program of the “Stichting voor Fundamenteel Onderzoek der Materie (FOM)”, which is financially supported by the “Nederlandse Organisatie voor Wetenschappelijk Onderzoek (NWO)”, and is part of an industrial partnership program between Philips and FOM. References

Figure 5. Average quantum efficiency obtained for different clusters of antennas (numbers indicating amount of antennas), vs the fraction of emitters coupled to the antennas. Line is guide to the eye. Insets: SEM images of antenna clusters, scale bar denotes 500 nm.

antennas with strongly interacting arms (i.e., with parameters L ) 80 nm and ∆ ) 20 nm) are shown in Figure 5 versus the estimated fraction of coupled molecules. The increase of the number of antennas in the cluster up to 4 further improves the quantum efficiency up to 59%. Extrapolation of our measurements to 100% coupled dye fraction points to a maximum achievable quantum efficiency of around 70%, in agreement with the calculated saturation due to nonradiative losses. In conclusion, we have observed resonant enhancement of light extraction from dye molecules by optical nanoantennas. By designing metal antennas with optimized arm length and gap width, the spontaneous emission of a fluorescent dye was enhanced by more than a factor of 5, while the quantum efficiency was improved from 40% to 53% with single nanoantennas. A clear effect of antenna coupling on the spontaneous emission enhancement was found and explained, using a theoretical model, by the strong local field enhancement in the antenna gap. We conclude that, although enhancement is limited by metal losses, these can be small enough to result in considerable quantum efficiency improvement. Acknowledgment. We acknowledge B. Ketelaars and P. Vergeer for technical assistance and O. T. A. Janssen, H. P. Urbach, W. Vos, H. Mertens, and A. Polman for stimulating discussions. V.G. and J.A.S.-G. acknowledge partial support from the Spanish Ministerio de Educacio´n y Ciencia (grants

Nano Lett., Vol. 7, No. 9, 2007

(1) Ozbay, E. Science 2006, 311, 189. (2) Barnes, W. L.; Dereux, A.; Ebbesen, T. W. Nature 2003, 424, 824. (3) Gontijo, I.; Boroditsky, M.; Yablonovitch, E.; Keller, S.; Mishra, U. K.; DenBaars, S. P.; Phys. ReV. B 1999, 60, 11564-11567. (4) Shimizu, K. T.; Woo, W. K.; Fisher, B. R.; Eisler, H. J.; Bawendi, M. G.; Phys. ReV. Lett. 2002, 89, 117401:1-4. (5) Kulakovich, O.; Strekal, N.; Yaroshevich, A.; Maskevich, S.; Gaponenko, S.; Nabiev, I.; Woggon, U.; Artemyev, M. Nano Lett. 2002, 2, 1449. (6) Anger, P.; Bharadwaj, P.; Novotny, L. Phys. ReV. Lett. 2006, 96, 113002. (7) Ku¨hn, S.; Hakanson, U.; Rogobete, L.; Sandoghdar, V. Phys. ReV. Lett. 2006, 97, 017402. (8) Dulkeith, E.; Ringler, M.; Klar, T. A.; Feldmann, J.; Munoz Javier, A.; Parak, W. J.; Nano. Lett. 2005, 5, 585-589. (9) Tam, F.; Goodrich, G. P.; Johnson, B. R.; Halas, N. J. Nano. Lett. 2007, 7, 496-501. (10) Biteen, J. S.; Lewis, N. S.; Atwater, H. A.; Mertens, H.; Polman, A. Appl. Phys. Lett. 2006, 88, 131109. (11) Gerber, S.; Reil, F.; Hohenester, U.; Schlagenhaufen, T.; Krenn, J. R.; Leitner, A.; Phys. ReV. B 2007, 75, 073404-1:4. (12) Schuck, P. J.; Fromm, D. P.; Sundaramurthy, A.; Kino, G. S.; Moerner, W. E. Phys. ReV. Lett. 2005, 94, 017402. (13) Mu¨hlschlegel, P.; Eisler, H.-J.; Martin, O. J. F.; Hecht, B.; Pohl, D. W. Science 2005, 308, 1607. (14) Farahani, J. N.; Pohl, D. W.; Eisler, H.-J.; Hecht, B. Phys. ReV. Lett. 2005, 95, 017402. (15) Taminiau, T. H.; Moerland, R. J.; Segerink, F. B.; Kuipers, L.; Van Hulst, N. F. Nano. Lett. 2007, 7, 28-33. (16) Aizpurua, J.; Bryant, G. W.; Richter, L. J.; Garcı´a de Abajo, F. J.; Kelley, B. K.; Mallouk, T. Phys. ReV. B 2005, 71, 235420. (17) Krenn, J. R.; Schider, G.; Rechberger, W.; Lamprecht, B.; Leitner, A.; Aussenegg, F. R.; Weeber, J. C. Appl. Phys. Lett. 2000, 77, 3379. (18) Imura, K.; Nagahara, T.; Okamoto, H. J. Phys. Chem. B 2004, 108, 16344. (19) Muskens O. L.; Giannini, V.; Sa´nchez-Gil, J. A.; Go´mez Rivas, J. 2006, cond-mat/0612689. (20) Lakowicz, J. R. Principles of Fluorescence Spectroscopy, 3rd ed.; Springer: New York, 2006. (21) Nordlander, P.; Oubre, C.; Prodan, E.; Li, K.; Stockman, M. I. Nano Lett. 2004, 4, 899. (22) (a) Ford, G. W.; Weber, W. H. Phys. Rep. 1984, 113, 195. (b) Gersten, J.; Nitzan, A. J. Chem. Phys. 1981, 75, 1139. (23) Sa´nchez-Gil, J. A.; Garcı´a-Ramos, J. V.; Me´ndez, E. R. Phys. ReV. B 2000, 62, 10515. (24) Novotny, L. Appl. Phys. Lett. 1996, 69, 3806.

NL0715847

2875